Right now, all tests are a bit old-school and under-optimised, sometimes using a bazooka to kill a fly. Running all the tests on real world data is nice but maybe too heavy.
- Same coverage could be achieved in a smarter way
- Do a real work on what's missing in the global coverage to try to test it (at least weekly)
- Add tests with synthetic data to speed up computation for on_push tests
- Write on disk behaviour is far from optimal
- Use more of pytest fixtures, etc.
- Any other brilliant ideas to speed up the tests and improve coverage
Right now, all tests are a bit old-school and under-optimised, sometimes using a bazooka to kill a fly. Running all the tests on real world data is nice but maybe too heavy.